The unique transcriptional response produced by concurrent estrogen and progesterone treatment in breast cancer cells results in upregulation of growth factor pathways and switching from a

In breast cancer, progesterone receptor (PR) positivity or abundance is positively associated with survival and treatment response. It was initially believed that PR was a useful diagnostic marker of estrogen receptor activity, but increasingly PR has been recognised to play an important biological role in breast homeostasis, carcinogenesis and metastasis.

Need et al BMC Cancer (2015) 15:791 DOI 10.1186/s12885-015-1819-3 RESEARCH ARTICLE Open Access The unique transcriptional response produced by concurrent estrogen and progesterone treatment in breast cancer cells results in upregulation of growth factor pathways and switching from a Luminal A to a Basal-like subtype Eleanor F Need1*, Luke A Selth2,3, Andrew P Trotta1,4, Damien A Leach1, Lauren Giorgio1, Melissa A O’Loughlin1, Eric Smith5, Peter G Gill6, Wendy V Ingman7,8, J Dinny Graham9 and Grant Buchanan1,3 Abstract Background: In breast cancer, progesterone receptor (PR) positivity or abundance is positively associated with survival and treatment response It was initially believed that PR was a useful diagnostic marker of estrogen receptor activity, but increasingly PR has been recognised to play an important biological role in breast homeostasis, carcinogenesis and metastasis Although PR expression is almost exclusively observed in estrogen receptor positive tumors, few studies have investigated the cellular mechanisms of PR action in the context of ongoing estrogen signalling Methods: In this study, we contrast PR function in estrogen pretreated ZR-75-1 breast cancer cells with vehicle treated ZR-75-1 and T-47D breast cancer cells using expression microarrays and chromatin immunoprecipitation-sequencing Results: Estrogen cotreatment caused a dramatic increase in the number of genes regulated by progesterone in ZR-75-1 cells In T-47D cells that have naturally high levels of PR, estrogen and progesterone cotreatment resulted in a reduction in the number of regulated genes in comparison to treatment with either hormone alone At a genome level, estrogen pretreatment of ZR-75-1 cells led to a 10-fold increase in the number of PR DNA binding sites detected using ChIP-sequencing Time course assessment of progesterone regulated genes in the context of estrogen pretreatment highlighted a series of important regulatory pathways, including those driven by epithelial growth factor receptor (EGFR) Importantly, progesterone applied to cells pretreated with estradiol resulted in switching of the PAM50-determined intrinsic breast cancer subtype from Luminal A to Basal-like, and increased the Oncotype DX® Unscaled Recurrence Score Conclusion: Estrogen pretreatment of breast cancer cells increases PR steady state levels, resulting in an unequivocal progesterone response that upregulates key members of growth factor pathways The transformative changes progesterone exerts on the breast cancer subtype suggest that these subtyping tools should be used with caution in premenopausal women Keywords: Progesterone receptor, Estrogen receptor, EGFR, Crosstalk, PAM50 * Correspondence: Eleanor.need@adelaide.edu.au Cancer Biology Group, The Basil Hetzel Institute for Translational Health Research, School of Medicine, The University of Adelaide, DX465701, 28 Woodville Road, Woodville South, 5011 South Australia, Australia Full list of author information is available at the end of the article © 2015 Need et al Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated Need et al BMC Cancer (2015) 15:791 Background Breast cancer is the most commonly diagnosed invasive cancer in females [1] and is most often an estrogen (17βestradiol) driven tumour [2, 3] The primary cellular mediator of estrogen is the intracellular transcription factor estrogen receptor alpha (ERα), which is expressed in 75 % of early breast cancers [4] ERα and PR positivity as assessed via immunohistochemistry of primary breast cancer is currently the gold standard indicator for hormonal therapy, applied either at the time of diagnosis or subsequent to surgical, chemotherapeutic and/or radiation management While the molecular mechanisms and consequences of estrogen-mediated action have received considerable research attention, the molecular mechanisms of progesterone signalling have not been as widely reported More recently PR is emerging as a key mediator of normal mammary gland development and tumorigenesis in mice, promoting mammary stem cell expansion and directing the immune microenvironment [5–10] The majority of the cellular effects of progesterone are mediated by the progesterone receptor (PR), an intracellular transcription factor of which two isoforms exist, PR-A and PR-B Because PR is an estrogen regulated gene, the expression of PR protein detected by immunohistochemistry as a diagnostic tool was found to discriminate between those most likely to respond to endocrine therapy, from those that will not [11, 12] Indeed, expression of PR in breast cancer in the absence of ERα is rare (1.5 % of cases), and evidence suggests that such cases may represent false negatives for ERα staining upon re-analysis [13–16] Nevertheless, PR appears to be more than a mere diagnostic indicator of estrogenic activity, as clinical studies have demonstrated it to be an independent biomarker of endocrine therapy response as well as a prognostic biomarker in postmenopausal breast cancers [12, 16–18] Smaller studies in premenopausal women have found that tumours containing higher PR positivity had the best response to tamoxifen [19] In premenopausal women, the physiological role of progesterone is inextricably linked to that of estrogen, with regards to production and secretion by the ovaries during the menstrual cycle Increased production of estrogen by the maturing follicles ultimately results in ovulation, after which the corpus luteum produces and secretes progesterone The secretion of progesterone in turn acts on the adrenal glands to stimulate a concomitant secondary, albeit smaller, peak of serum estrogen [20] Evidence also suggests that the postmenopausal breast is capable of sequestering and/or synthesising progesterone and estrogen from circulating hormonal precursors [21–25] Collectively, it appears most likely that PR is activated within a hormonal milieu that includes active estrogen signalling Genomic and functional studies of receptor action in vitro now provide unprecedented detail into the precise Page of 17 mechanics of ERα and, to a lesser extent, PR action in breast cancer cells Those for PR have, however, been exclusively performed in the absence of exogenous estrogen [26–31] Binding of estrogen by ERα and progesterone by PR results in association of the receptors with specific sites on chromatin Receptor binding to DNA subsequently directs the recruitment of cofactors and associated coactivators and corepressors, resulting in modification of the local chromatin landscape and activation or repression of target genes Indirect tethering of the receptors to chromatin has also been observed via interaction with DNA-bound factors such as AP-1, Stat3 and SP1 [27, 32, 33] Despite the findings that PR expression is almost always accompanied by ERα expression [16], to date there are few reported studies investigating progesterone transcriptional signalling and PR binding in the context of estrogen-mediated signalling Indeed, most studies of PR DNA binding have been performed in T-47D breast cancer cells that not depend upon estrogen for PR expression [34] In this report, we demonstrate a 10-fold induction in PR binding upon progesterone treatment in estrogen pre-treated versus non estrogen treated ZR-75-1 cells and demonstrate that progesterone and estrogen cotreatment drive a unique gene expression profile in ZR-75-1 that is distinct from treatment with either hormone alone, which includes up-regulation of signalling mediators of ErbB pathways Estrogen and progesterone cotreatment cause significant changes to the predicted intrinsic breast cancer subtype, specifically to one that resembles more aggressive, therapy resistant disease Methods Cell lines and culture ZR-75-1, T-47D, MCF-7, MDA-MB-231, BT-20 and MDA-MB-453 cells were obtained from the American Type Culture Collection (Rockville, MD) and maintained in RPMI 1640 (Life Technologies, NSW, Australia) containing 10 % (ZR-75-1) or % (T-47D, MCF-7) fetal bovine serum (FBS) (Sigma-Aldrich, NSW, Australia) All experiments were performed within 20 passages of supply from ATCC (Manassas, Virginia) Immunoblot analysis ZR-75-1, T-47D, MCF-7, MDA-MB-231, BT-20 and MDA-MB-453 cells were seeded in well plates at × 105 cells/well in phenol red free RPMI 1640 containing to 10 % hormone stripped FBS (SigmaAldrich), in the proportions indicated for each cell type above Hormone stripped treatment medium was supplemented with 10nM estrogen where indicated After 72 h, medium was replaced with the indicated hormone treatment for the specified time Cells were lysed, protein concentration assessed, electrophoresed Need et al BMC Cancer (2015) 15:791 and transferred to Hybond-C membrane as previously described [31] Membranes were probed using AR-N20, PRH190, ERα-HC20, CTSD-H75, FKBP5-H100 (Santa Cruz Biotechnology, CA), calnexin (CANX, Thermo Scientific, VIC, Australia), and anti-tubulin alpha (TUBA, Millipore, VIC, Australia) and detected as previously described [31] Microarray, RNA extraction and RT-qPCR Cells were plated for 72 h in 6-well plates in phenol redfree RPMI 1640 containing 10 % hormone stripped FBS at x 105/ well, treated for 16 h with vehicle (ethanol; V.C), 10nM estrogen, 10 nM progesterone, 10 nM estrogen + 10nM progesterone, or for 72 h with 10nM estrogen (pretreated) with or without subsequent 10nM progesterone for 4, or 16 h RNA was extracted using RNeasy kit (Qiagen, VIC, Australia) The ZR-75-1 microarray results presented in Fig represent findings from quadruplicate samples randomly hybridised to Illumina HumanWG-6v3 chips (Australian Genome Research Facility, St Lucia, Australia) Raw transcript expression data was exported from Illumina BeadStudio software and analysed using the Bioconductor Limma package implemented in R [35], as previously described [31] Briefly, we normalised array data using variance stabilisation normalisation [36], corrected the data with Combat [37], filtered to likely expressed transcripts (~24,000) and subjected the data to linear model fitting Regulation compared to vehicle was accepted for an empirical Bayes moderated t-statistic incorporating Benjamini-Hochberg correction of ≤0.05 Microarrays in T-47D cells presented in Fig were performed in triplicate and were hybridised to Illumina HumanWG-6v2 chips (Genomics Core, Norris Comprehensive Cancer Centre, University of Southern California, USA) Raw transcript expression data was processed as described above, but subjected to two Combat corrections due to array batch effects Samples for the ZR-75-1 time course microarray presented in Fig were generated in × 105 cells per well in well plates in triplicate from ZR-75-1 cells treated with 72 h 10nM estrogen or vehicle, followed by 4, or 16 h 10nM progesterone treatment Hormone treatments were performed by overlaying the progesterone treatment on the existing media and the experiment was performed with reverse timing so all samples were collected at the same time point Triplicate RNA samples were hybridised to human Gene 1.0 ST Affymetrix Arrays (Adelaide Microarray Centre, Adelaide, Australia) Raw CEL files were normalised, filtered for expressed transcripts (~23,875) and subjected to linear model fitting Regulation compared to E2 pretreated samples was accepted for P4 treated samples for a Bayes moderated tstatistic with Benjamini-Hochberg correction of ≤0.0001, yielding a total of 2140 genes regulated at some point over the whole time course Validation for all microarray Page of 17 results was performed on independent RNA samples by RT-qPCR using iQ SYBR Green Supermix (BioRad Life Science, NSW, Australia) on the CFX-96 PCR machine (Bio-Rad) Primer sequences are provided in Additional file All microarray data is available online at NCBI (accessions GSE61538, GSE61368 and GSE62243) Pathway overrepresentation analysis was performed on differentially expressed genes using the comprehensive, publicly available InnateDB database, with hypergeometric testing and Benjamini-Hochberg correction for false discovery rates [38] Clustering of microarray data was performed using the K-means clustering method, with 20 random starts in STEM, and a maximum output set to model profiles [39] Cell cycle studies ZR-75-1 cells were plated in well plates in phenol redfree RPMI 1640 containing 10 % hormone stripped FBS and 10nM estrogen at × 105/ well for 72 h Cells were then treated with 10nM progesterone or equivalent vehicle for 24 h Cells were washed in PBS, harvested and fixed in ice cold 70 % ethanol Fixed cells were incubated in 50 μg/ml propidium iodide (Sigma Aldrich), 40 μg/ml RNAse A (Life Technologies, NSW, Australia) and 0.1 % Tween20 (Sigma Aldrich) in PBS for h in the dark Cell cycle analysis was conducted on a FACSCanto II running DIVA software (BD Bioscience, NSW, Australia) DNA frequency histograms were obtained using FlowJo software (Treestar, Oregon, USA) using the Dean-Jett-Fox model Results are representative of three independent experiments Chromatin immunoprecipitation (ChIP) and ChIP-sequencing ChIP and ChIP-sequencing was performed as previously described [31] Briefly, ZR-75-1 and T-47D cells were plated for 72 h in phenol red-free RPMI 1640 containing 10 % hormone stripped FBS with 10nM estrogen or equivalent vehicle After 72 h, medium was supplemented with the indicated hormone for h Immunoprecipitation was performed with PR-H190X or normal rabbit IgG antisera (Santa Cruz Biotechnology, CA) In total, independent ChIP experiments were performed, each independently validated by RT-qPCR at an enhancer region of FKBP5 and a nonspecific DNA region Peaks were called and analysis was performed as described in [31] Briefly, Genomic regions with a peak height of (minimum of independent 36 bp reads/site on a Illumina Genome Analyser II) were recorded using FindPeaks4 (Vancouver Short Read Analysis Package; http://vancouvershortr.sourceforge.net/) on human genome build 18 (hg18) and subsequent analysis was performed in R using custom algorithms as outlined in [31] Bed files are provided as Additional files, and the primary data has been deposited at NCBI Manipulation of Need et al BMC Cancer (2015) 15:791 Fig (See legend on next page.) Page of 17 Need et al BMC Cancer (2015) 15:791 Page of 17 (See figure on previous page.) Fig Estrogen and progesterone induce a unique transcriptomic response in ZR-75-1 and T-47D cells a Protein steady state levels of ERα, PR-A, PR-B, androgen receptor (AR), androgen and progesterone regulated gene FKBP5 and estrogen regulated gene CTSD in ZR-75-1, T-47D and MCF-7 cells treated with ethanol (v.c.), 10nM DHT, 10nM PROG or 10nM estrogen for 16 h TUBA and calnexin (CANX) were utilised as controls Note that exposure time was different for each cell line and was optimised to visualise changes in response to hormone treatment b Non hormone treated protein steady state levels of ERα, PR-A and PR-B in ZR-75-1, T-47D and MCF-7 cells treated with v.c for 16 h Alpha tubulin (TUBA) was utilised as a control Exposure times were different from the blot presented in Fig 1a c Microarray analysis of the transcriptomic response of ZR-75-1 cells treated with ethanol (v.c.), 10 nM estrogen, 10 nM PROG, or cotreated with 10 nM estrogen and 10 nM PROG for 16 h Euler diagram (left) demonstrates commonly regulated genes and those uniquely regulated by the hormonal cotreatment Histograms (right) demonstrate validation of progesterone-regulated responses in independent samples Expression presented relative to housekeeping gene GAPDH expression (d) Microarray analysis of the transcriptomic response of T-47D cells treated with ethanol (v.c.), 10 nM estrogen, 10 nM PROG, or cotreated with 10 nM estrogen and 10 nM PROG for 16 h Euler diagram (left) demonstrates commonly regulated genes in response to each treatment Histograms (right) demonstrate validation of progesterone-regulated responses in independent samples e Cell cycle analysis of propidium iodide stained ZR-75-1 cells after treatment for 24 h with vehicle (V.C; ethanol), 10nM progesterone or pretreated for 72 h with 10nM estrogen (E2p), followed by 16 or 24 h of 10nM progesterone treatment (E2p + P4) intervals for analysing overlaps between different PR ChIP-seq datasets was performed in R, Galaxy [40] or BiSA [41] The ChIP-seq datasets Conservation of binding sites amongst vertebrates was performed using the Cistrome Analysis Pipeline (http://cistrome.dfci.harvard.edu/ ap) Regions of PR binding were annotated with respect to neighbouring genes using ChIPpeakAnno [42] and CisGenome [43] High confidence sites were defined by our ability to empirically validate selected PR binding sites in independent samples (Additional file 2) To compare strength of PR binding at specific peak subsets, sequence tag libraries were generated and average tag density at the subsets was determined using the peak annotation function in HOMER v4.2 [44] Novel sequence motifs that were present in PR binding regions statistically significantly more frequently than expected by random chance were identified using Gibbs Motif Sampling [45] or MEME [46] Known sequence motifs in the JASPAR CORE vertebrata database [47] that were significantly enriched in the PR cistrome were identified using CisGenome, with default parameters [47, 48] Fold enrichment and significance (Fisher’s exact test) of motif sequences were estimated compared with an equal number of 1-kb control regions with matched physical distribution Results Shaping of the progesterone response by estrogen in breast cancer cells To ascertain the most appropriate breast cancer cell line model to investigate the physiological progesterone response in the context of estrogen signalling, we assessed alterations in steady state protein levels of ERα, PR, androgen receptor (AR), Cathepsin D (CTSD) and FK506 binding protein (FKBP5) in response to estrogen, progesterone and 5α-dihydrotestosterone (DHT) in a panel of breast cancer cell lines Of the cell lines tested, only MCF7, T-47D and ZR-75-1 had detectable levels of both ERα and PR upon immunoblotting (Fig 1a and Additional file 3) As the results in Fig 1a were obtained with different exposure times, depending on the steady state level of the protein, we then compared the relative steady state levels of ERα and PR in MCF7, T-47D and ZR-75-1 cells and found that ZR-75-1 cells had the most equivalent detectable expression of all three receptors (Fig 1b) Upon estrogen treatment, increased steady state levels of PR and CTSD were most dramatic in ZR-75-1 and T-47D cells, indicating activation of ERα We observed that treatment of the cell lines with progesterone resulted in increased steady state levels of FKBP5 in T-47D cells but not in ZR-75-1 cells (Fig 1a) This observation is not due to methodological artefacts as we were able to observe an increase in FKBP5 in ZR-75-1 cells in response to the androgen 5alpha-dihydrotestosterone (DHT) To examine the potential regulatory effects of progesterone in the presence and absence of estrogen signalling, we performed microarray expression profiling of ZR-75-1 and T-47D cells following treatment with vehicle, estrogen, progesterone or both ligands in combination Only genes were regulated by progesterone alone in ZR-75-1 cells (SERPINA3 and SEPT4; see Additional file 4) In contrast to these results, we were able to observe a small but consistent increase in FKBP5 expression upon RT-qPCR in ZR-75-1 cells in response to progesterone treatment, which was not detected using our cutoff criteria for differential expression on microarray (Fig 1c; Benjamini-Hochberg corrected Bayesian moderated t-statistic p < 0.05) In agreement, this small increase in expression did not result in increased FKBP5 steady state levels upon progesterone treatment as observed by immunoblotting (Fig 1c versus Fig 1a) In contrast to the minimal effect of progesterone alone in ZR-75-1 cells, cotreatment with estrogen and progesterone resulted in significant regulation of 216 genes (Benjamini-Hochberg corrected Bayesian moderated t-statistic p < 0.05; Fig 1c; see Additional file 4) Although 170 of these genes were also regulated upon estrogen treatment alone (78.7 %; Fig 1c; see Additional file 4), 46 (21.3 %) were unique to the progesterone and estrogen cotreatment In addition, cotreatment with progesterone resulted in the loss of regulation of 56 genes Need et al BMC Cancer (2015) 15:791 (25 %) observed with estrogen treatment alone (Fig 1c; see Additional file 4) In T-47D cells in contrast, treatment with progesterone alone resulted in regulation of 329 genes, of which 87 (26 %) were also significantly regulated by estrogen alone (Fig 1d; Additional file 5) Estrogen and progesterone cotreatment resulted in the loss of regulation of 24.9 % of estrogen responsive genes and 19.8 % of progesterone responsive genes In contrast to ZR-75-1, only genes were uniquely responsive to estrogen and progesterone cotreatment in T-47D cells (GJB2, SSBP1 and ZFP36), and far fewer were regulated upon estrogen and progesterone cotreatment; 79 in T-47D, 216 in ZR-75-1 (Compare Fig 1c to d) Results using independent sets of RNA samples reflect those findings, with candidate genes (FKBP5, THOC5, SERPINA3) showing significant upregulation in response to estrogen and progesterone cotreatment in ZR-75-1 cells, but no effect of estrogen and progesterone cotreatment in T-47D on these candidates in comparison to progesterone treatment alone (Fig 1c) When the transcriptomic profiles of ZR-75-1 cells cotreated with progesterone plus estrogen were compared with T-47D treated with either progesterone only or estrogen plus progesterone, only 9.8 % (21/214) and 11 % (25/214) of genes were found to be in common Collectively, these data indicate that the cotreatment of ZR-75-1 cells with estrogen sensitises the cells to progesterone and produces a unique transcriptional response that is distinct from the response mediated by estrogen or progesterone alone in either ZR75-1 or T-47D cells Pathway analysis was performed separately on progesterone upregulated and down regulated genes in T-47D cells Both of the gene lists were enriched for genes involved in cell cycle In the upregulated gene list, transcriptional pathways were enriched, and pathways involved in DNA synthesis were significantly enriched in the downregulated gene list (see Additional file 6A and B) In estrogen and progesterone cotreated T-47D cells, fewer genes were regulated, but hormonal actions were over represented, such as glucocorticoid receptor regulation (see Additional file 6C and D) Enrichment of hormonal pathways was more evident in estrogen and progesterone treated ZR-75-1 cells, along with enrichment of genes involved in growth factor receptor signalling (Additional file 7A and B) These results suggest that estrogen and progesterone cotreatment in ZR-75-1 and T-47D cells produces a different transcriptomic response from progesterone alone in either cell type Hence, the physiological effect of estrogen pretreatment on ZR-75-1 responsiveness to progesterone was assessed via cell cycle analysis using flow cytometry Administration of progesterone to ZR-75-1 cells pretreated for 72 h with estrogen resulted in an small increase in the proportion of cells in the replicative S and G2M phases of the cell cycle, and fewer in the quiescent G0-G1 phases Page of 17 (Fig 1e) This effect was not observed in cells treated with progesterone only and is consistent with those previously observed in other breast cancer cell lines and with the in vivo response in mice to estrogen and progesterone cotreatment [49, 50] Estrogen pretreatment increases PR genomic occupancy To characterise PR action in the context of estrogen treatment, we performed PR ChIP-seq in ZR-75-1 cells treated with progesterone alone or after estrogen pretreatment of the cells with 72 h of 10nM estrogen DNA pooled from independently validated ChIP experiments (Additional file 8) was subjected to nextgeneration sequencing After adjusting for input (see methods), 49,927 progesterone alone and 75,030 estrogen pretreated + progesterone binding sites were scored Using these data, we identified 475 high confidence binding sites in the progesterone alone PR cistrome and 4597 high confidence estrogen pretreated + progesterone binding sites (Additional file 9; sites in bed format) Only 31 of those high confidence sites were shared between the two cistromes, and had a much greater average peak height in comparison to sites not shared between the cistromes (Additional file 10A) Parallel analysis in T-47D cells validated these as likely PR binding sites, but there was little evidence of increased enrichment upon estrogen pretreatment (Additional file 10B) Western blotting revealed increased PR steady state levels in ZR-75-1 cells following estrogen pretreatment (Fig 2a) The estrogen pretreated and progesterone alone PR binding sites are unique Comparison of putative PR binding sites revealed a much greater sequence conservation amongst vertebrates for the progesterone treated, estrogen pretreated binding sites than the binding sites identified after treatment with progesterone alone, as well as a greater number of reads per peak (Fig 2b-d) Using Gibbs Motif Sampling and MEME analysis approaches, the most highly enriched de novo motif in the estrogen pretreated PR cistrome resembled canonical PR binding sites, which were over-represented 3.24 and 3.69 respectively in comparison to the background genome average (Fig 2e; p =